File size: 7,454 Bytes
42fb3af
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
"""
NetworkX DiGraph construction from extracted entities + trials.
Upserts nodes and edges (merges PMIDs, recomputes confidence average).
Pre-populates with ALS seed entities from config.
"""
from __future__ import annotations

import json
from pathlib import Path

import networkx as nx

from config import (
    ALS_SEED_ENTITIES,
    ENTITIES_PATH,
    KG_MIN_EDGE_CONFIDENCE,
    TRIALS_PATH,
)
from extraction.normalizer import normalize_entity
from logging_config import get_logger
from models import PaperExtractionResult

_logger = get_logger("graph.builder")


def build_graph(
    entities_path: Path = ENTITIES_PATH,
    trials_path: Path = TRIALS_PATH,
) -> nx.DiGraph:
    """Build the ALS knowledge graph. Returns a populated DiGraph."""
    G: nx.DiGraph = nx.DiGraph()

    _add_seed_entities(G)
    _logger.info(f"Seeded graph: {G.number_of_nodes()} seed nodes")

    if entities_path.exists():
        n_papers = _add_extracted_entities(G, entities_path)
        _logger.info(
            f"After extraction: {G.number_of_nodes()} nodes, "
            f"{G.number_of_edges()} edges from {n_papers} papers"
        )
    else:
        _logger.warning(f"Entities file not found: {entities_path} — skipping NER enrichment")

    if trials_path.exists():
        n_trials = _add_trials(G, trials_path)
        _logger.info(f"Added {n_trials} trial nodes")
    else:
        _logger.warning(f"Trials file not found: {trials_path}")

    return G


def _add_seed_entities(G: nx.DiGraph) -> None:
    type_map = {
        "genes": "Gene",
        "proteins": "Protein",
        "compounds": "Compound",
        "mechanisms": "Mechanism",
        "phenotypes": "Phenotype",
    }
    for category, entity_type in type_map.items():
        for name in ALS_SEED_ENTITIES.get(category, []):
            canonical_id = normalize_entity(name, entity_type)
            _upsert_node(G, canonical_id, {
                "type": entity_type,
                "display_name": name,
                "paper_count": 0,
                "evidence_pmids": [],
                "is_seed": True,
            })


def _add_extracted_entities(G: nx.DiGraph, entities_path: Path) -> int:
    n_papers = 0
    with open(entities_path, encoding="utf-8") as f:
        for line in f:
            line = line.strip()
            if not line:
                continue
            raw = json.loads(line)
            pmid = raw["pmid"]
            n_papers += 1

            # Add / update entity nodes
            for ent in raw.get("entities", []):
                canonical_id = ent.get("canonical_id", "")
                if not canonical_id:
                    continue
                if G.has_node(canonical_id):
                    G.nodes[canonical_id]["paper_count"] += 1
                    G.nodes[canonical_id]["evidence_pmids"].append(pmid)
                    # Update confidence as running average
                    cur = G.nodes[canonical_id].get("confidence", 0.7)
                    G.nodes[canonical_id]["confidence"] = (cur + ent.get("confidence", 0.7)) / 2
                else:
                    _upsert_node(G, canonical_id, {
                        "type": ent.get("type", "Unknown"),
                        "display_name": ent.get("name", canonical_id),
                        "paper_count": 1,
                        "evidence_pmids": [pmid],
                        "confidence": ent.get("confidence", 0.7),
                        "is_seed": False,
                    })

            # Add / update relationship edges
            for rel in raw.get("relationships", []):
                source = rel.get("source", "")
                target = rel.get("target", "")
                rel_type = rel.get("relation_type", "")
                if not source or not target or not rel_type:
                    continue

                # Ensure both endpoints exist as nodes
                for node_id in (source, target):
                    if not G.has_node(node_id):
                        _upsert_node(G, node_id, {
                            "type": "Unknown",
                            "display_name": node_id.split(":", 1)[-1],
                            "paper_count": 0,
                            "evidence_pmids": [],
                            "is_seed": False,
                        })

                conf = rel.get("confidence", 0.7)
                if G.has_edge(source, target):
                    edge = G[source][target]
                    if rel_type not in edge.get("relation_types", []):
                        edge.setdefault("relation_types", [edge.get("relation_type", rel_type)])
                        edge["relation_types"].append(rel_type)
                    if pmid not in edge["evidence_pmids"]:
                        edge["evidence_pmids"].append(pmid)
                    # Running average confidence
                    edge["confidence"] = (edge["confidence"] + conf) / 2
                else:
                    G.add_edge(source, target, **{
                        "relation_type": rel_type,
                        "relation_types": [rel_type],
                        "evidence_pmids": [pmid],
                        "confidence": conf,
                        "evidence_text": rel.get("evidence_text", ""),
                    })

    return n_papers


def _add_trials(G: nx.DiGraph, trials_path: Path) -> int:
    n_trials = 0
    with open(trials_path, encoding="utf-8") as f:
        for line in f:
            line = line.strip()
            if not line:
                continue
            trial = json.loads(line)
            nct_id = trial.get("nct_id", "")
            if not nct_id:
                continue

            node_id = f"trial:{nct_id}"
            _upsert_node(G, node_id, {
                "type": "ClinicalTrial",
                "display_name": trial.get("title", nct_id)[:120],
                "nct_id": nct_id,
                "phase": trial.get("phase", ""),
                "status": trial.get("status", ""),
                "url": trial.get("url", f"https://clinicaltrials.gov/study/{nct_id}"),
                "paper_count": 0,
                "evidence_pmids": [],
            })

            # Link trial to known target entities
            for target_name in trial.get("target_entities", []):
                # Try gene, compound, protein
                matched = False
                for etype in ("Gene", "Compound", "Protein", "Mechanism"):
                    candidate_id = normalize_entity(target_name, etype)
                    if G.has_node(candidate_id):
                        if not G.has_edge(node_id, candidate_id):
                            G.add_edge(node_id, candidate_id, **{
                                "relation_type": "TESTED_IN",
                                "relation_types": ["TESTED_IN"],
                                "evidence_pmids": [],
                                "confidence": 1.0,
                                "evidence_text": "",
                            })
                        matched = True
                        break
                if not matched:
                    _logger.debug(f"Trial {nct_id}: no graph node for target {target_name!r}")

            n_trials += 1
    return n_trials


def _upsert_node(G: nx.DiGraph, node_id: str, attrs: dict) -> None:
    if G.has_node(node_id):
        G.nodes[node_id].update(attrs)
    else:
        G.add_node(node_id, **attrs)